Training-Free Indoor White Space Exploration

Exploration of white spaces has been recognized as a promising way to improve the utilization of wireless spectrums. Especially, indoor white space exploration has been shown to be much more challenging than that in outdoor environment. Most of existing works on indoor white space exploration infer availabilities of TV channels based on the correlations among the channels and locations, which is learned from the training data measured beforehand. However, the process of training data collection normally requires considerable time and devices, as well as human power. In this paper, we perform a measurement of indoor white spaces and study their characteristics. Based on the in-depth understanding of the characteristics, we propose a Training-free Indoor white space exploration MEchanism (TIME). In TIME, we design an algorithm for channel state inference based on Bayesian compressive sensing, as well as an incremental method for deployment of spectrum detectors. Furthermore, we present an algorithm to determine a proper number of spectrum detectors in need. Extensive real-world experiments are conducted to evaluate TIME's performance. The evaluation results demonstrate that TIME achieves competitive performances against the state-of-the-art training-based mechanisms.